The role of indirect plant-plant interactions via
shared pollinators: a combined experimental and
theoretical study in species-rich temperate
grasslands.
Dissertation
der Mathematisch-Naturwissenschaftlichen Fakultät der Eberhard Karls Universität Tübingen
zur Erlangung des Grades eines Doktors der Naturwissenschaften
(Dr. rer. nat.)
vorgelegt von Sven Hanoteaux aus Chimay, Belgien
Tübingen
Tag der mündlichen Qualifikation: 11 July 2014
Dekan: Prof. Dr. Wolfgang Rosenstiel
1. Berichterstatter: Prof. Dr. Katja Tielbörger
Citation
Hanoteaux, S. 2014. The role of indirect plant-plant interactions via
shared pollinators: a combined experimental and theoretical study in
species-rich temperate grasslands. Ph.D. Thesis. University of Tübingen, Germany.
Publication
During this thesis, the following article has been published :
Hanoteaux, S., Tielbörger, K. & Seifan, M. (2013). Eects of spatial patterns on the pollination success of a less attractive species. Oikos, 122, 867880.
Declaration of my own contribution to the present manuscript
The programming of the model used in the rst chapter was entirely my work. The eld experiment was jointly designed with Eva-Maria Hoch. I designed the common garden experiment and gathered all the data concerning pollinator visits for the two last chapters. The statistical data analysis of the three chapters was entirely my work.
During the course of the thesis I was advised by Dr. Merav Seifan, with whom I discussed the results. Dr. Merav Seifan and Prof. Dr. Katja Tielbörger contributed to advanced drafts of the rst chapter, the published manuscript, as co-authors. The other two chapters were entirely written by me, but proofread by Dr. Merav Seifan and Prof. Dr. Katja Tielbörger.
Contents
Acknowledgements 9
Abstract 11
Synopsis 13
Chapter 1
Eects of spatial patterns on the pollination success of a less attractive
species 27
Chapter 2
Density and spatial distribution of an attractive species alter plant-pollinator
interaction structure in grasslands - A network approach. 65
Chapter 3
Identity of neighbouring species alters the response of pollinators to oral
density in articial plant communities 105
Discussion 133
Acknowledgements
Long was the way, I would not have made it alone...
First of all, I'm very grateful to Dr. Merav Seifan for having introduced me to the eld of pollination ecology, for her guidance and support, constructive criticisms on my work, for always nding the time to discuss my results or improving my manuscripts, for her encouragements and, yes also for Sandman. Thank you so much Merav!
I also thank Prof. Dr. Katja Tielbörger for giving me the chance to write my PhD in her working group, for the prime times she spent correcting my manuscripts, for her support during the time in Tübingen and especially to have triggered that reaction when motivation was gone. I would not have made it without it. Thank you Katja!
I owe a deep debt to Sara Bangerter, who translated my application form into German and as such allowed me to get funded for two years. Thank you Sara!
Evidently, I also would like to thank the Land Baden-Württenberg for funding through the Landesgraduiertenförderung Baden-Württemberg fellowship.
During the time in Tübingen, I met many people who were the source of inspiring thoughts. I would like to thank Pierre for teaching to me what ecology is about and that it can (should?) be done drinking coee and smoking; Peter who thaught me much about modelling with just one question; Tal for always nding the time to answer my little questions and for spending hours answering them, greatly improving my under-standing of programming; Mark for his help in many aspects of my PhD, Wolfgang for
having made my life much easier in terms of computer use and for helping me spreading R through the world; Mira, not only for the nicest summer ever in the eld but also for Johnossi; Betty and Astrid for teaching to me the nicest word in German: Notstro-maggregat; Marcos for teaching to me how to make my work attractive; Ortrun for her invaluable help in setting up experiments and raising plants; Andrea and Christiane for making German paperwork look easy and Michael for teaching me the names of plants. For the help received in dierent ways, I thank Sabine, Srijana, Christian, Johannes, Tom, Philipp, Nicole, Jan, Mirka, Hongchun, Michal, Udi, Yolanda, Anne, Benedikt, Raúl, Nina, Sara, Sara and Sara.
Not necessarily involved in the scientic aspects of my PhD, but certainly involved in the social aspects of it, I would like to thank Thierry, Kim, Paco, Ines, Ingo, Mark, Kris, Niklas, Jonas, Bertrand, Antoine, Fix & Emina & PJ, Sébastien, Nicolas, Jean-Pascal, Bert, Johannes, Germain, Laurent, Denis and Fabrice.
I also would like to thank all persons involved in the development of free operating systems and softwares for making such a great job!
Thank you so much Maman, Armand, Malika and Danaë, I guess you know why. A special thanks to Ingo - it all started because of you!
Abstract
Anthropogenic activities are rapidly changing the world. The ongoing climate change (and its associated shifts in owering phenologies), biological invasions and increased fragmentation of ecosystems are all inducing rapid changes in structural characteristics of plant communities. In communities, where many species depend on the service of pollinators for their reproduction, changes in species composition, oral densities and spatial distribution will undoubtedly further aect plant-pollinator interactions.
Due to the interdependence of plant and pollinator species for their reproduction, plant-pollinator interactions are central in the maintenance of both plant and plant-pollinator species and hence of biodiversity in many ecosystems. Therefore, understanding how structural characteristics of plant communities are aecting plant-pollinator interactions, would en-able a better anticipation of the ecological consequences of destructive human activities. The present thesis investigated the impact of changing structural characteristics of Euro-pean grassland plant communities on plant-pollinator interactions and on the outcome of indirect plant-plant interactions mediated through shared pollinators. Firstly, a spatially explicit model was used to theoretically examine the interplay between the densities and the spatial distribution of two dierently attractive species on the plant species survival. Secondly, the results of the model were put to the test in a eld experiment in which the changes in plant-pollinator structure induced by changes in densities and spatial patterns were tracked using a network approach. Finally, a common garden experiment allowed us to test the importance of species identities on the role they play on the outcome of indirect plant-plant interactions.
The overall ndings clearly indicate that the small-scale spatial patterning in plant com-munities is an important factor shaping the outcome of indirect plant-plant interactions
by manipulating the behaviour of pollinators. Spatial aggregation in plant species can result in pollinators being trapped in mono-specic patches, substantially increasing the quality and the quantity of pollinator services received by the aggregated plant species. This spatial mechanism is especially strong when the aggregated species grows at high density. This mechanism was shown to aect not only the survival of a less attractive species in a theoretical model but also the structure of plant-pollinator interactions un-der natural situation, by altering patterns of resource use by pollinators. Hence, spatial distribution of plant species at small-scale and its impact on the pollinator behaviour should be considered as an important process in shaping the general characteristics of plant-pollinator networks.
Further, the identities of plant species and their associated set of oral traits are un-doubtedly inuencing choices made by foraging pollinators and hence, shape the outcome of indirect plant-plant interactions via shared pollinators. Our results showed that the outcome of these indirect interactions is likely to be dictated by the dierence in attrac-tiveness among species rather than by the attractive character of only one species. The occurrence and intensity of both intra- and interspecic density dependent responses in the pollinator behaviour were shown to be conditioned not only by the identity of a focal species but also by the identity of its neighbours.
Future research should try to incorporate the small-scale spatial distribution of species and a characterisation of the attractiveness dierential among owering plant species in the set of ecological factors important in shaping the outcome of indirect plant-plant interactions. This would enable a better anticipation of the impact of anthropogenic changes on plant-pollinator interactions and by extension on plant reproductive success and as such on patterns of species coexistence shaping the structure of plant communities.
Synopsis
Introduction
A vast majority of plant species rely on the services of pollinators for their reproduction (Ollerton et al. 2012). Further, pollinator species are as well dependent on the oral reward oered by visited plant species for their survival and reproduction (Westrich 1989). As such, plant-pollinator interactions are crucial to the maintenance of both plant and pollinator species, and hence of biodiversity in many ecosystems (Geber & Moeller 2006, Waser & Ollerton 2006, Mitchell et al. 2009), a central question in ecology (see e.g. Loreau et al. 2001 for a recent review). Moreover, the insight that most pollinator species are generalists in their use of oral resources (Waser et al. 1996, Waser & Ollerton 2006) induces a strong potential for indirect plant-plant interactions mediated through shared pollinators (Rathcke 1983, Waser & Ollerton 2006, Sargent & Ackerly 2008, Hegland et al. 2009). In diverse plant communities dependent on the services of pollinators for their reproduction, these indirect interactions may play an important role in shaping patterns of species coexistence (Sargent & Ackerly 2008, Mitchell et al. 2009). The outcome of such interactions can either be positive (i.e. facilitation, Rathcke 1983, Laverty 1992, Johnson et al. 2003, Feldman et al. 2004, Moeller 2004, Moragues & Traveset 2005, Ghazoul 2006, Bartomeus et al. 2008, Muñoz & Cavieres 2008, Hegland et al. 2009), negative (i.e. competition, Rathcke 1983, Grabas & Laverty 1999, Moragues & Traveset 2005, Bartomeus et al. 2008, Hegland et al. 2009) or neutral (no interactions, Grabas & Laverty 1999, Moragues & Traveset 2005, Muñoz & Cavieres 2008, Hegland et al. 2009). The major determinant of the nature of these indirect interactions is to be sought in
the way pollinators will respond to what they perceive while foraging and the subsequent choices made (Kunin & Iwasa 1996, Chittka & Thomson 2001, Feldman et al. 2004, Lãzaro & Totland 2010). Hence, plant communities structural characteristics such as the identity, density and spatial distribution of species in community are all likely to determine the nature of the eects co-owering species exert on each other (Rathcke 1983, Grabas & Laverty 1999, Moragues & Traveset 2005, Muñoz & Cavieres 2008, Lãzaro & Totland 2010) as these factors are shaping the structure of the visual and olfactory landscape of foraging pollinators.
One of the rst attempts to articulate a model predicting the outcome of these indirect plant-plant interactions mediated through shared pollinators, is the visitation density relationship developed by Rathcke (1983). This graphical model states that a low oral density, visitation rate is very low and any small increase in density, created by either the same species' individuals or by another species which shares pollinators, will have a positive eect on the visitation rate and thus also on the reproductive success. However, as the number of pollinators available at a given place and time is nite, the outcome of such indirect interactions would shift from facilitation to competition with further increase in oral density (Ratchke 1983). Another well-known example of (positive) indirect plant-plant interactions, is the so called magnet species eect (Laverty 1992, Johnson et al. 2003, Moeller 2004). By locally increasing pollinator abundance, an attractive, highly rewarding species can increase the pollination success of rewardless or less attractive neighbouring species (e.g. Laverty 1992, Johnson et al. 2003, Molina-Montenegro et al. 2008). However, the close proximity with a highly attractive species could also result in strong competition if pollinators focus their foraging eorts more on the attractive species (Chittka & Schürkens 2001, Muñoz & Cavieres 2008) or if the close vicinity of the attractive species increases improper pollen transfer (e.g. Brown et al. 2002, Cariveau. & Norton 2009).
However, choices made by pollinators while foraging in patches are not only dened by the oral density and attractiveness of species (Chittka & Thomson 2001). Even if some pollinators species are known to travel substantial distances to nd resources patch (Os-borne et al. 1999, Pasquet et al. 2008, Hagen et al. 2011), they tend to y short distances between consecutive visits (Waser 1982) and have restricted maximum detection ranges of visual and olfactory cues (Ne'eman & Kevan 2001). This implies that foraging bouts in patches are conducted within a restricted spatial extent. Hence, the spatial patterning of species can also aect the behaviour of pollinators (Goulson 1994, Morales & Vazquez 2008), by manipulating their foraging landscape. For example, a spatially clumped species might oer to a visiting pollinator a locally increased availability of that species in each clump (Goulson 1994, Feldman et al. 2004). This will tend to increase the visit
quan-tity (Rathcke 1983) in these clumps and increase the visit quality (by reducing improper pollen transfer rates, Rathcke 1983). The inverse is expected when species are spatially well mixed in communities, as pollinators will perceive a more heterogeneous foraging landscape, potentially inducing a switching behaviour and as such a decrease in visitation quality (Rathcke 1983, Brown et al. 2002). The potential eects of the spatial patterning of species within community on the reproductive success of plant species becomes even more complex when considering that they also depend on the ability of pollinators to dis-cern among species (Chittka & Thomson 2001), on their innate preferences (Giurfa et al. 1995, Chittka et al. 1999, Chittka & Thomson 2001, Raine & Chittka 2007, Ings et al. 2009) and on change in preferences of pollinators with recent foraging experience (Dukas & Real 1993, Keasar et al. 1996) or oral resource availability (Goulson 1994, Kunin & Iwasa 1996, Chittka & Thomson 2001).
Even if often acknowledged, the occurrence and intensity of such spatial processes were never thoroughly investigated in pollination ecology (Goulson 1994, Feldman et al. 2004). In order to ll this gap, we used a spatially explicit individual based model to investigate the role played by relative densities and spatial patterning on the survival of two species diering only in their attractiveness to pollinators. Unlike many previous models (Bobisud & Neuhaus 1975, Waser 1978, Goulson 1994, Feldman et al. 2004), the model incorporated an elaborate pollinator behaviour which included innate preferences, the ability to dis-criminate among species and changes in preferences according to the foraging experience, reecting hence a more realistic pollinator behaviour.
In species rich natural communities such as temperate grasslands, the impact on polli-nation patterns of the two above mentioned factors will be far more complex to predict than when considering only two species (Dunne et al. 2002, Bascompte et al. 2003, Olesen et al. 2007, Hegland et al. 2009). In such communities, plant-pollinator interactions are forming highly complex networks of interactions, the so-called pollination networks (Jor-dano 1987, Dunne et al. 2002, Bascompte et al. 2003, Blüthgen et al. 2008, Olesen et al. 2011). Due to the high interconnection between the two trophic levels in these networks, changes in the plant community composition and structure and the subsequent adapta-tion in the pollinator behaviour will propagate along the many network connecadapta-tions and can substantially aect the plant reproductive success of species in these communities (Bascompte et al. 2003, Waser & Ollerton 2006, Blüthgen et al. 2008, Olesen et al. 2011). The use of networks in pollination ecology is relatively new (Jordano 1987) but the study of general characteristics of pollination networks has known a rapid development (Jordano 1987, Dunne et al. 2002, Bascompte et al. 2003, Blüthgen et al. 2006, Olesen et al. 2007, Blüthgen et al. 2008, Dormann et al. 2009, Olesen et al. 2011). The development of this branch of pollination ecology is fortunate as it allows to tackle highly complex problems
with a relatively simple methodological approach. As such, we now have a large set of network describers available (the so-called network metrics, Dormann et al. 2008, 2009) allowing not only the characterisation of general network properties (Bascompte et al. 2003, Vazquez & Aizen 2004, Olesen et al. 2007) but also allowing to track changes in pollination patterns induced by changes in structural characteristics of plant communities (Lopezaraiza-Mikel et al. 2007, Tylianakis et al. 2007, Bartomeus et al. 2008, Morales & Vazquez 2008) and to relate them to the occurrence of biological processes (Santamaría & Rodríguez-Gironéz 2007, Blüthgen et al. 2008, Vazquez et al. 2009).
In the second part of the present thesis, a network approach was used in order to evaluate the ecological consequences of introducing a highly attractive plant species in a species rich grassland community. By manipulating the density and the spatial patterns of the introduced species, we aimed at investigating changes in the network structure induced by our experimental manipulations and hence put the results of the model developed in the rst chapter, to the test. We focussed on the analysis of ecological relevant indices related to the general organisation of interactions within these networks, to diversity and evenness of the interactions, to patterns of resource use by pollinators and nally to specialisation in pollinator behaviour.
Further, in the results of studies investigating the outcome of indirect plant-plant inter-actions, all types of interactions were found: positive, negative or neutral (see above for relevant references). This discrepancy in the results of such studies reects that even though we have acquired a good understanding of the possible mechanisms shaping these indirect interactions (Rathcke 1983, Laverty 1992, Feldman et al. 2004, Seifan et al. 2014), we are still unable to predict their outcome. The eects of that a species exerts on its neighbours are undoubtedly species specic as it will depend on both its oral traits and the cognitive ecology of the visiting pollinators. As such, it seems logical to think that an attractive species will be more likely to aect its neighbours than a less attractive species (Laverty 1992). However, the concept attractiveness itself is relative and is only dened in a given ecological context. A species' attractiveness will not only be determined by its oral traits but also by the identity (and hence oral traits) of its neighbours, i.e. the attractiveness dierential and not the absolute attractiveness will be important. This may explain why the above mentioned results are not conclusive.
Hence in the last part of this thesis, we aimed at investigating how changes in the species identities would aect the behavioural responses of pollinators to changing relative densi-ties of two plant species. We conducted a common garden experiment in which we created articial communities composed of two species, along a gradient of their relative densities. By systematically exchanging the identity of the species in our communities, we aimed at investigating the role played by the oral traits of both a focal species and the oral
traits of its dierent neighbours.
In summary, the present study aimed at disentangling the eects the oral density, the spatial patterns and the identity of plant species on the outcome of indirect plant-plant interactions mediated through shared pollinators in species rich grasslands. This thesis presents a unique combination of theoretical, observational and experimental approaches whose aim was to investigate largely understudied aspects of pollination ecology. The importance of this work goes beyond the sole investigation of the impact of unexplored structural community characteristics on the outcome of indirect plant-plant interactions. Indeed, the recent decline in pollinator populations and diversity (Kearns et al. 1998, Biesmeijer et al. 2006, Potts et al. 2010) could have severe impacts on these interac-tions and cascades of extincinterac-tions are to be expected if the trend in species loss continues (Waser & Ollerton 2006, Olesen et al. 2007). Additionally to species loss, anthropogenic changes such as biological invasions (Traveset & Richardson 2006, Morales & Traveset 2009), landscape fragmentation (see Aguilar et al. 2006 for a recent review) and shifts in owering phenologies due to global warming (Visser & Both 2005, Elzinga et al. 2007, Miller-Rushing et al. 2010, González-Varo et al. 2013), have brought substantial changes in structural characteristics of plant communities. Furthermore changes in species iden-tities, densities and spatial distribution of species in communities undoubtedly aect the foraging experience of pollinators and hence their behaviour (Schemske 1981, Rathcke 1983, Dukas & Real 1993, Keasar et al. 1996, Chittka & Thomson 2001, Lãzaro & Tot-land 2010). Alterations in the pollinator behaviour will have repercussions on the plant-pollinator interactions and by extension on both plant and plant-pollinator species reproduction and survival. Hence, it is capital to understand how pollinators adapt their behaviour to the dierent structural characteristics of plant communities in order to be able to better anticipate further anthropogenic changes.
Thesis objectives and organisation
The present thesis is organised in three distinct chapters. The overall aim of this thesis was to investigate the eects of structural plant community characteristics on the outcome of indirect plant-plant interactions through shared pollinators. Hence, each chapter reects a separate investigation of the combination of the oral density with such a structural aspect. Explicitly, the objectives of the individual chapters were as follows.
Chapter 1: This chapter aimed at theoretically investigating the impact of spatial patterning and its interplay with relative oral density on the reproductive success of
two species diering only in their attractiveness for pollinators. This was done using a spatially explicit individual based model coupled with an agent based model allowing the modelling of a complex pollinator behaviour.
Chapter 2: This chapter aimed at assessing changes in the structure of plant-pollinator interactions induced by the introduction of an attractive species in semi-natural grass-lands, via the analysis of pollination networks. The impacts of the spatial distribution and the density of the attractive species were tested by analysing a large set of network indices related to ecologically relevant aspects of the structure of plant-pollinator interactions. Chapter 3: This chapter aimed at testing whether the density responses in the be-haviour of pollinators foraging in articial two species plant communities is altered by changes in plant species identities. Systematically changing the identities of the two species allowed us to investigate the eects of both species on both intra- an interspecic density dependence in the pollinator behavioural responses.
Key Results
Chapter 1:
Eects of spatial patterns on the pollination success of a less
at-tractive species
By using a spatially explicit individual based model coupled with an agent based model allowing the modelling of a complex pollinator behaviour, we evaluated the eects of the relative densities and spatial distribution of two species dierently attractive to
pollinators on the survival of the less attractive species.
We found that, at low relative density, the less attractive species had a higher survival when spatially uniformly distributed than when spatially aggregated. On the other hand, when the less attractive species was more abundant (i.e. at high relative density), its sur-vival was higher when spatially aggregated in mono-specic patches than when uniformly distributed. These results indicate that spatially aggregated species can trap pollinators in mono-specic patches. These results were consistent as long as the scale of the plant spatial aggregation was similar to or larger than the pollinators' detection range. Finally, a certain degree of generalisation in the pollinator behaviour was the necessary condition for the eect of spatial patterns to emerge.
Chapter 2:
Density and spatial distribution of an attractive species alter
plant-pollinator interaction structure in grasslands.
In order to put the predictions of the theoretical model developed in the rst chapter of this thesis, to the test, we introduced an attractive species into semi-natural grasslands and manipulated its density and spatial distribution in a full factorial way. A large set of network indices reecting important ecological processes in plant-pollinator
interactions were analysed in order to track the changes in network structure following the introduction of the attractive species and the manipulation of its density and spatial distribution.
Our results suggest that the neutrality hypothesis can explain the changes in diversity and evenness of plant-pollinator interactions following the introduction of the attractive species and its density manipulation. However, it fails to explain the observed impact of spatial patterns. Indeed, we found that a regular spacing of the attractive species induced a higher exclusivity and lower similarity in resource use by pollinators than when the attractive species was spatially clumped. We have proven that small-scale spatial mechanisms are at work in pollination patterns and conrmed the predictions of the previous chapter. Our results further suggest that the introduced species act as a strong competitor for the services of pollinators, especially at high density.
Chapter3:
Identity of neighbouring species alters the response of pollinators
to oral density in articial plant communities
We created articial communities composed of two dierently attractive species and built a gradient of their relative oral densities. By systematically exchanging the two species, we aimed at investigating the eects of their identities (i.e. of both the focal species and of its neighbour) on both intra- and interspecic density dependent responses in the behaviour of several important pollinator groups.
We found that the identity of neighbouring species can induce and/or alter both intra-and interspecic density dependent response in quantitative aspects of the pollinator behaviour. Both positive and negative interspecic interactions among plant species were found and always enabled by the main shared pollinators and conditioned by the identity of the neighbouring species. Changing the identity of the neighbouring species can hence
alter the outcome of indirect plant-plant interactions. These results were explained in the light of the dierence in attractiveness among plant species for pollinator, leading to the brief introduction of the concept of attractiveness dierential.
References
Aguilar, R., Ashworth, L., Galetto, L. & Aizen, M.A. (2006). Plant reproductive suscepti-bility to habitat fragmentation: review and synthesis through a meta-analysis. Ecology Letters, 9, 968980.
Bartomeus, I., Vila, M. & Santamaria, L. (2008). Contrasting eects of invasive plants in plant-pollinator networks. Oecologia, 155, 761770.
Bascompte, J., Jordano, P., Melian, C.J. & Olesen, J.M. (2003). The nested assembly of plant-animal mutualistic networks. Proceedings of the National Academy of Sciences, 100, 93839387.
Biesmeijer, J.C., Roberts, S.P.M., Reemer, M., Ohlemuller, R., Edwards, M., Peeters, T., Schaers, A.P., Potts, S.G., Kleukers, R., Thomas, C.D., Settele, J. & Kunin, W.E. (2006). Parallel declines in pollinators and insect-pollinated plants in Britain and the Netherlands. Science, 313, 351354.
Blüthgen, N., Frund, J., Vazquez, D.P. & Menzel, F. (2008). What do interaction network metrics tell us about specialization and biological traits? Ecology, 89, 33873399. Blüthgen, N., Menzel, F. & Blüthgen, N. (2006). Measuring specialization in species
interaction networks. BMC Ecology, 6, 9.
Bobisud, L.E. & Neuhaus, R.J. (1975). Pollinator constancy and survival of rare species. Oecologia, 21, 263272.
Brown, B., Mitchell, R. & Graham, S. (2002). Competition for pollination between an invasive species (purple loosestrife) and a native congener. Ecology, 83, 23282336. Cariveau., D.P. & Norton, A.P. (2009). Spatially contingent interactions between an
exotic and native plant mediated through ower visitors. Oikos, 118, 107114.
Chittka, L. & Schürkens, S. (2001). Successful invasion of a oral market - An exotic Asian plant has moved in on Europe's river-banks by bribing pollinators. Nature, 411, 653.
Chittka, L. & Thomson, J.D. (2001). Cognitive ecology of pollination : animal behaviour and oral evolution. Cambridge University Press.
Chittka, L., Thomson, J.D. & Waser, N.M. (1999). Flower constancy, insect psychology, and plant evolution. Naturwissenschaften, 86, 361377.
Dormann, C.F., Gruber, B. & Fruend, J. (2008). Introducing the bipartite package: Analysing ecological networks. R News, 8, 811.
Dormann, C.F., Jochen, F., Nico, B. & Bernd, G. (2009). Indices, graphs and null models: Analyzing bipartite ecological networks. The Open Ecology Journal, 2, 724.
Dukas, R. & Real, L.A. (1993). Eects of recent experience on foraging decisions by bumble bees. Oecologia, 94, 244246.
Dunne, J.A., Williams, R.J. & Martinez, N.D. (2002). Food-web structure and network theory: The role of connectance and size. Proceedings of the National Academy of Sciences, 99, 1291712922.
Elzinga, J.A., Atlan, A., Biere, A., Gigord, L., Weis, A.E. & Bernasconi, G. (2007). Time after time: owering phenology and biotic interactions. Trends in Ecology & Evolution, 22, 432439.
Feldman, T., Morris, W. & Wilson, W. (2004). When can two plant species facilitate each other's pollination? Oikos, 105, 197207.
Geber, M.A. & Moeller, D.A. (2006). The ecology and evolution of owers, Oxford Uni-versity Press, Oxford, UK, chap. Pollinator responses to plant communities and impli-cations for reproductive character evolution, pp. 102119.
Ghazoul, J. (2006). Floral diversity and the facilitation of pollination. Journal of Ecology, 94, 295304.
Giurfa, M., Nunez, J., Chittka, L. & Menzel, R. (1995). Colour preferences of ower-naive honeybees. Journal of Comparative Physiology A: Neuroethology, Sensory, Neural, and Behavioral Physiology, 177, 247259.
González-Varo, J.P., Biesmeijer, J.C., Bommarco, R., Potts, S.G., Schweiger, O., Smith, H.G., Stean-Dewenter, I., Szentgyörgyi, H., Woyciechowski, M. & Vilà, M. (2013). Combined eects of global change pressures on animal-mediated pollination. Trends in Ecology & Evolution, 28, 524530.
Goulson, D. (1994). A model to predict the inuence of insect ower constancy on inter-specic competition between insect pollinated plants. Journal of Theoretical Biology, 168, 309314.
Grabas, G. & Laverty, T. (1999). The eect of purple loosestrife (Lythrum salicaria L,; Lythraceae) on the pollination and reproductive success of sympatric co-owering wetland plants. Ecoscience, 6, 230242.
Hagen, M., Wikelski, M. & Kissling, W.D. (2011). Space use of bumblebees (Bombus spp.) revealed by radio-tracking. PLoS ONE, 6, e19997.
Hegland, S.J., Grytnes, J.A. & Totland, Ø. (2009). The relative importance of positive and negative interactions for pollinator attraction in a plant community. Ecological Research, 24, 929936.
Ings, T., Raine, N. & Chittka, L. (2009). A population comparison of the strength and persistence of innate colour preference and learning speed in the bumblebee bombus terrestris. Behavioral Ecology and Sociobiology, 63, 12071218.
Johnson, S., Peter, C., Nilsson, L. & Agren, J. (2003). Pollination success in a deceptive orchid is enhanced by co-occurring rewarding magnet plants. Ecology, 84, 29192927. Jordano, P. (1987). Patterns of mutualistic interactions in pollination and seed dispersal:
Connectance, dependence asymmetries, and coevolution. American Naturalist, 129, 657677.
Kearns, C.A., Inouye, D.W. & Waser, N.M. (1998). Endangered mutualisms: The con-servation of plant-pollinator interactions. Annual Review of Ecology and Systematics, 29, 83112.
Keasar, T., Shmida, A. & Motro, U. (1996). Innate movement rules in foraging bees: Flight distances are aected by recent rewards and are correlated with choice of ower type. Behavioral Ecology and Sociobiology, 39, 381388.
Kunin, W. & Iwasa, Y. (1996). Pollinator foraging strategies in mixed oral arrays: Density eects and oral constancy. Theoretical Population Biology, 49, 232263. Laverty, T.M. (1992). Plant interactions for pollinators visits: a test of the magnet species
eect. Oecologia, 89, 502508.
Lopezaraiza-Mikel, M.E., Hayes, R.B., Whalley, M.R. & Memmott, J. (2007). The im-pact of an alien plant on a native plant-pollinator network: an experimental approach. Ecology Letters, 10, 539550.
Loreau, M., Naeem, S., Inchausti, P., Bengtsson, J., Grime, J.P., Hector, A., Hooper, D.U., Huston, M.A., Raaelli, D., Schmid, B., Tilman, D. & Wardle, D.A. (2001). Bio-diversity and ecosystem functioning: Current knowledge and future challenges. Science, 294, 804808.
Lãzaro, A. & Totland, Ø. (2010). Local oral composition and the behaviour of pollinators: attraction to and foraging within experimental patches. Ecological Entomology, 35, 652 661.
Miller-Rushing, A.J., Høye, T.T., Inouye, D.W. & Post, E. (2010). The eects of pheno-logical mismatches on demography. Philosophical Transactions of the Royal Society B: Biological Sciences, 365, 31773186.
Mitchell, R.J., Flanagan, R.J., Brown, B.J., Waser, N.M. & Karron, J.D. (2009). New frontiers in competition for pollination. Annals of Botany, 103, 14031413.
Moeller, D.A. (2004). Facilitative interactions among plants via shared pollinators. Ecol-ogy, 85, 32893301.
Molina-Montenegro, M.A., Badano, E.I. & Cavieres, L.A. (2008). Positive interactions among plant species for pollinator service: assessing the `magnet species' concept with invasive species. Oikos, 117, 18331839.
Moragues, E. & Traveset, A. (2005). Eect of Carpobrotus spp. on the pollination success of native plant species of the Balearic islands. Biological Conservation, 122, 611619. Morales, C.L. & Traveset, A. (2009). A meta-analysis of impacts of alien vs. native plants
on pollinator visitation and reproductive success of co-owering native plants. Ecology Letters, 12, 716728.
Morales, J.M. & Vazquez, D.P. (2008). The eect of space in plant - animal mutualistic networks: insights from a simulation study. Oikos, 117, 13621370.
Muñoz, A.A. & Cavieres, L.A. (2008). The presence of a showy invasive plant disrupts pollinator service and reproductive output in native alpine species only at high densities. Journal of Ecology, 96, 459467.
Ne'eman, G. & Kevan, P. (2001). The eect of shape parameters on maximal detection distance of model targets by honeybee workers. Journal Of Comparative Physiology A-neuroethology Sensory Neural and Behavioral Physiology, 187, 653660.
Olesen, J.M., Bascompte, J., Dupont, Y.L., Elberling, H., Rasmussen, C. & Jordano, P. (2011). Missing and forbidden links in mutualistic networks. Proceedings of The Royal Society B-Biological Sciences, 278, 725732.
Olesen, J.M., Bascompte, J., Dupont, Y.L. & Jordano, P. (2007). The modularity of pollination networks. Proceedings of the National Academy of Sciences, 104, 19891 19896.
Ollerton, J., Price, V., Armbruster, W.S., Memmott, J., Watts, S., Waser, N.M., Ørjan Totland, Goulson, D., Alarcon, R., Stout, J.C. & Tarrant, S. (2012). Overplaying the role of honey bees as pollinators: a comment on Aebi and Neumann (2011). Trends in Ecology & Evolution, 27, 141142.
Osborne, J., Clark, S., Morris, R., Williams, I., Riley, J., Smith, A., Reynolds, D. & Ed-wards, A. (1999). A landscape-scale study of bumble bee foraging range and constancy, using harmonic radar. Journal of Applied Ecology, 36, 519533.
Pasquet, R.S., Peltier, A., Huord, M.B., Oudin, E., Saulnier, J., Paul, L., Knudsen, J.T., Herren, H.R. & Gepts, P. (2008). Long-distance pollen ow assessment through evaluation of pollinator foraging range suggests transgene escape distances. Proceedings of the National Academy of Sciences, 105, 1345613461.
Potts, S.G., Biesmeijer, J.C., Kremen, C., Neumann, P., Schweiger, O. & Kunin, W.E. (2010). Global pollinator declines: trends, impacts and drivers. Trends in Ecology & Evolution, 25, 345353.
Raine, N.E. & Chittka, L. (2007). The adaptive signicance of sensory bias in a foraging context: Floral colour preferences in the bumblebee bombus terrestris. PLoS ONE, 2, e556.
Rathcke, B. (1983). Pollination biology, Academic Press, Orlando, Florida, chap. Com-petition and facilitation among plants for pollination, pp. 305329.
Santamaría, L. & Rodríguez-Gironéz, M. (2007). Linkage Rules for Plant-Pollinator Net-works: Trait Complementarity or Exploitation Barriers? PLoS ONE, 5, e31.
Sargent, R.D. & Ackerly, D.D. (2008). Plant-pollinator interactions and the assembly of plant communities. Trends in Ecology & Evolution, 23, 123130.
Schemske, D.W. (1981). Floral convergence and pollinator sharing in 2 bee-pollinated tropical herbs. Ecology, 62, 946954.
Seifan, M., Hoch, E.M., Hanoteaux, S. & Tielbörger, K. (2014). Shifts in the outcome of shared pollination services are aected by the density and spatial pattern of an attractive neighbor. Accepted in Journal of Ecology.
Traveset, A. & Richardson, D.M. (2006). Biological invasions as disruptors of plant re-productive mutualisms. Trends in Ecology & Evolution, 21, 208216.
Tylianakis, J.M., Tscharntke, T. & Lewis, O.T. (2007). Habitat modication alters the structure of tropical host-parasitoid food webs. Nature, 445, 202205.
Vazquez, D.P. & Aizen, M.A. (2004). Asymmetric specialization: a pervasive feature of plant-pollinator interactions. Ecology, 85, 12511257.
Vazquez, D.P., Chaco, N.P. & Cagnolo, L. (2009). Evaluating multiple determinants of the structure of plant-animal mutualistic networks. Ecology, 90, 20392046.
Visser, M.E. & Both, C. (2005). Shifts in phenology due to global climate change: the need for a yardstick. Proceedings of the Royal Society B: Biological Sciences, 272, 25612569. Waser, N.M. (1978). Interspecic pollen transfer and competition between co-occurring
plant species. Oecologia, 36, 223236.
Waser, N.M. (1982). A comparison of distances own by dierent visitors to owers of the same species. Oecologia, 55, 251257.
Waser, N.M., Chittka, L., Price, M.V., Williams, N.M. & Ollerton, J. (1996). General-ization in pollination systems, and why it matters. Ecology, 77, 10431060.
Waser, N.M. & Ollerton, J. (2006). Plant-pollinator interactions: from specialization to generalization. University Press. Chicago.
Westrich, P. (1989). Die Wildbienen Baden-Wuerttembergs. vol. I and II. Ulmer Verlag, Stuttgart.
CHAPTER
1
Eects of spatial patterns on the pollination success of a less
attractive species
Abstract
Plant individuals rely on pollinator services for their reproduction and often have to share these services with co-occurring neighbours, creating complex indirect plant-plant inter-actions. Many current theoretical models focus on the eect of oral resources' density on the net outcome of these indirect plant-plant interactions, often neglecting the identity of plant species in the communities and especially the species' spatial pattern. To ll this gap, we created a spatially explicit model whose goal was to study the interplay between relative densities and spatial distribution patterns of two plant species diering in their attractiveness for pollinators. Since theory predicts that pollinator behaviour strongly governs the outcome of pollination, we allowed the pollinators to systematically change their plant preferences based on their foraging experience. Thus the interplay between density and spatial pattern of plants was tested over a continuum of behaviours from spe-cialists to generalists. Our most striking nding was that reproductive success of the less attractive was aected in an opposite way by spatial patterns depending on whether the species had relatively low or high densities. Namely, when the less attractive species was highly abundant, its survival was higher when aggregated in large monospecic patches than when uniformly distributed. On the other hand, when the attractive species was more abundant, the less attractive species survived better when uniformly distributed.
These results were consistent as long as the scale of the plant spatial aggregation was similar to or larger than the pollinators' detection range. Our results suggest that ag-gregated plant spatial patterns manipulate pollinator behaviour by trapping them within monospecic patches. This eect was suciently strong to enhance the survival of a com-petitively inferior species and hence to act in a way similar to the more familiar niche or temporal separation among plant species.
Introduction
Reproduction of many plant species is subject to pollination success and often pollinator services are shared among co-owering species (Campbell & Motten 1985, Feinsinger 1987, Geber & Moeller 2006, Mitchell et al. 2009). Therefore, community composition and the identity of neighbouring plants are likely to aect individual reproductive success. For example, the preferences and foraging patterns of pollinators (and hence plant reproduc-tive success) are not merely the outcome of species-specic oral traits, but are greatly aected by the oral composition of the entire plant community (Kunin 1997, Chittka & Thomson 2001, Ghazoul 2006, Lãzaro & Totland 2010). In particular, pollinator foraging patterns may be strongly aected by the identity of the neighbouring plants because of dierential attractiveness of oral display (Clegg & Durbin 2000), variation in reward content and quality (Dukas & Real 1993a, Klinkhamer & van der Lugt 2004), or because of inherent preferences and foraging behaviour of dierent pollinator groups (Sih & Baltus 1987, Lãzaro & Totland 2010). The intensity by which plants aect reproductive success of their neighbours is obviously aected by their relative attractiveness, i.e. the impact of attractive species on 'unattractive' ones is most likely much larger than vice-versa. There-fore, a useful approach to study the impact of neighbourhood community structure on the outcome of shared pollinator services is to focus on less attractive species, and determine their reproductive success as a function of modied features of their neighbouring plant species.
The factors enhancing the success of less attractive species in a community with attractive plants may be theoretically classied into two groups: the rst is composed of factors which act against the negative eect of attractive neighbours. Since attractive plants are, by denition, preferred by pollinators, any factor that reduces the pollinator's ability to choose among species and forces it to visit the less attractive species will increase the less attractive species's reproductive success. One such factor is the relative density of the species in the community (Sih & Baltus 1987). If a species's density is relatively high, visitation rate may be increased simply due to the functional and/or numerical
response of pollinators, irrespective of its attractiveness (see e.g. Klinkhamer & van der Lugt 2004). The second group of factors is connected to the potential ability of the more attractive species to facilitate reproductive success of its neighbours. In these cases, the attractive species serves as a magnet species, increasing local pollinator activity and thus actively increasing not only its own reproductive success but also that of its less attractive neighbours (Laverty 1992, Johnson et al. 2003, Moeller 2004, Juillet et al. 2007).
Previous studies attempted to generate a robust theoretical background for predicting the outcome of both processes and hence to unravel the prevailing factors acting when plants share pollinators. One of the earliest models predicted that pollinator visitation rate per ower will increase with increasing plant density until pollinator visits are saturated and competition for pollinator visits starts dominating (Rathcke 1983). At low densities, visitation rate is very low and any small increase in density, created by either the same species' individuals or by another species which shares pollinators, will have a positive eect on the visitation rate and thus will increase reproductive success. However, as density increases further, competition for pollinators will become increasingly important. Kunin & Iwasa (1996) found similar results, but showed that the relative disadvantage of the low density species can be reduced by a manipulation of the pollinator's foraging choices, namely a specialisation of some pollinators on the low density species. A positive eect of a heterogeneous species composition at low plant densities was supported by an analytical model by Feldman et al. (2004) which showed that if the pollinator visitation rate is an initially accelerating function of total ower density, plant species showed higher reproductive success and longer survival time in the presence of another species relative to monocultures.
Next to the identity and density of neighbours, an important but largely understudied factor that may determine reproductive success of unattractive species is spatial arrange-ment of plant individuals. For example, within a plant community, a spatially clumped distribution of a certain species may oer a locally increased resource availability within each clump of that particular species. This will tend to increase pollinator visits in these clumps (increased visit quantity, Rathcke 1983), and in addition reduces the negative ef-fects of improper pollen transfer (increased visit quality, Rathcke 1983). When the species are spatially well-mixed within a community, this tendency is expected to be reversed be-cause the pollinator may perceive a more heterogeneous foraging landscape which may reduce visitation quality. Therefore, spatial segregation of oral resources could be ben-ecial for species due to lower interspecic competition (Goulson 1994, Jakobsson et al. 2009), similar to the positive eects of intraspecic clumping related to other resources (Stoll & Prati 2001).
considering that they depend also on pollinator traits, i.e. their ability to dierentiate between plant species and to respond to the perceived vegetation patterns. Unfortunately, many previous models used relatively simple rules for pollinator foraging behaviour (Bo-bisud & Neuhaus 1975, Waser 1978, Goulson 1994, Feldman et al. 2004), and there is a lack of studies combining the recipient and provisioning part of pollination in determining plant community structure. This discrepancy between the two aspects of pollination ecol-ogy is unfortunate, because there are clear indications that pollinator decision-making plays a signicant role in the outcome of such systems (Kunin & Iwasa 1996, Chittka & Thomson 2001). For example, pollinators are known to y shorter distances between two consecutive successful visits (Waser 1982) which reduces the spatial extent of the foraging bout. This observation, coupled with restricted maximum detection ranges of visual and olfactory cues (Ne'eman & Kevan 2001), indicates the importance of plant community composition and structure. Furthermore, many pollinator groups have cer-tain innate preferences for owering traits such as colours (Giurfa et al. 1995, Chittka et al. 1999, Chittka & Thomson 2001, Raine & Chittka 2007, Ings et al. 2009). However, these preferences can change during foraging activity depending on the availability of o-ral resources (Goulson 1994, Chittka & Thomson 2001) and recent foraging experiences (Dukas & Real 1993b, Keasar et al. 1996). This implies that constancy may change ac-cording to the relative density and the identity of species found during a foraging bout (Grindeland et al. 2005, Hegland & Totland 2005, Cariveau. & Norton 2009). Plant com-munity spatial distribution can therefore alter pollinator behaviour (and thus constancy) by changing the available (i.e. detectable) resources, inducing dierent foraging experi-ences and hence aecting pollination success. Taking these considerations into account, the reproductive success of species should depend on their attractiveness, abundance and spatial distribution relative to the other species in the community.
In this study, we used an individual based simulation model to study the eect of plant community spatial pattern and pollinator characteristics on the probability of a less at-tractive species to survive. We employed a model where pollinators were foraging in a plant community composed of two species with dierent attractiveness to the pollinators. We varied the size of monospecic patches within the community and the relative densities of the two plant species in order to conduct an analysis of the interplay between density and spatial patterns. Since we assume that pollinator decision-making plays an impor-tant role, we tested the eect of three components: pollinator constancy (i.e. generalist vs. specialist behaviour), reward variation among plants, and pollinator detection range (Field Of View). Finally, we conducted a sensitivity analysis to test for the eects of dierent pollinator population sizes and plant dispersal distance on the model outcomes. We used this combined approach (i.e. considering pollinators and plant individuals
explic-itly) in order to test the following hypotheses: (i) in a community dominated by attractive species, less attractive individuals will be avoided if they grow in clumps, due to the high availability of preferred oral resources. However, (ii) if the less attractive species is randomly dispersed among attractive ones, pollinators will be more homogeneously dis-tributed in space, increasing the number of chance visits to the less attractive species. On the other hand, (iii) if the community is mainly composed of less attractive individuals, a spatial aggregation of the less attractive species may manipulate the pollinators' foraging landscape by reducing their choices over large areas of the plant community. This should increase the reproductive success of the less attractive species relative to a random distri-bution, where the attractive species can be detected and visited by pollinators from any position in the plant community.
Model
In order to investigate the eects of spatial patterns on the survival probability of a less attractive species, we developed a spatially explicit and time discrete model with a two-species plant community and a pollinator population. We used a grid based Individual Based Model (IBM) to model two self-incompatible owering plant species. In this model, plant species reproductive success, and thus their ability to survive in the community, was governed by pollinator behaviour. To achieve that, pollinators were modelled as agents which interacted with plant individuals during their foraging bouts. The spatial aspects of the vegetation model combined with an interactive model, explicitly considering plants and pollinators separately but interacting with each other, allowed us to test the inuence of spatial characteristics of both plants and their pollinators (initial spatial distribution of plant species and spatial movement of pollinators) in addition to the eects of non-spatial characteristics (e.g. densities of plant species, number of pollinators, oral constancy between visits).
The model contained a grid of 100 Ö 100 cells. Each cell represented a site in which a single plant can establish, ower, reproduce and die. We let the plants and pollinator interact with each other for 50 years and used the number of cells occupied by each plant species as a measure for its success. To avoid edge eects, the grid was designed as a torus. In the following, we briey describe our model assumptions concerning plant species traits, pollinator traits, and the interactions among the two trophic levels.
Plant community
Plant species
The plant community was composed of two annual plant species, which dier only in their attractiveness to pollinators (A; A ∈{0.1 , 0.9} ; where higher A values indicate higher
attractiveness; see Appendix 3 for the motivation for choosing attractiveness values). At the beginning of each model generation (i.e. one plant year), plant individuals of the two species were introduced to the grid as adult owering individuals. Each individual ower, regardless of its identity, could contain reward with a probability PR. As a default, both
plant species had a reward probability of 0.5 (i.e. pollinator landing on a ower had 50% chances to be rewarded). However, as part of the model sensitivity test, we also studied situations in which both species had dierent reward probabilities (see section Simulation experiments). For simplicity, plants did not rell their reward during a generation. If a plant individual was pollinated, it producedNSeed seeds that were then dispersed among
the surrounding grid cells at the end of the generation. Dispersal distance was based on a normal distribution with mean 0 and standard deviation ddisp. The direction of
dispersal was dened as an angle drawn from a uniform distribution ([0-360°[) (details see Appendix 4). At the end of each generation, all plant individuals died and the grid occupancy for the next generation was determined, taking into account the relative seed number of each species (weighted lottery; Warner & Chesson 1985; Appendix 4). For simplicity, no further competitive interactions between species were incorporated in the model (Straw 1972, Bobisud & Neuhaus 1975). Each plant individual could also die before reproducing with a probability PDeath = 0.05 and each cell had a probability PEmpty =
0.05 to stay empty during a generation. Community structure: spatial patterns
To systematically study the eects of spatial patterns on the outcome of plant-pollinator interactions, we manipulated the level of intraspecic spatial aggregation of plant com-munities. An example may be seen in Figure 1. The rst pattern generated was a ran-dom pattern (S1), where each cell was ranran-domly assigned to one of the two species with 0.5 probability. This random pattern can be seen as the random distribution of square monospecic patches with a one cell edge. To create more aggregated spatial patterns, we increased intra-specic aggregation by randomly assigning a square monospecic patch with an edge length of ve (S5) or ten (S10) cells (i.e. patches of 25 and 100 conspecic individuals, respectively) to one of the two species. As a control we also generated a regular arrangement of the two species among the grid (Reg; see details below).
Figure 1: Example of initial spatial patterns and relative species densities combinations used in this study. Black cells represent areas occupied by the attractive species and white cells represent areas occupied by the less attractive species. For simplicity, empty cells are not represented although they were present with a probability ofPEmpty< 0.05.
Community structure: plant density
Because previous studied indicated that species density may have a strong eect on repro-ductive success, we tested the spatial patterns with three dierent population densities, dened by the relative abundance of the attractive species (D = 10, 50 or 90 percent
of the non-empty cells were occupied by the attractive species). To create intraspecic aggregated patterns (S1, S5, and S10) with dierent species' densities, we divided the grid into the appropriate number of square patches which were then assigned randomly to a species according to its density: 10%, 50% or 90% of the patches were assigned to the attractive species, respectively. The remaining patches were then lled by the less attractive species. During this process, we took into account that a cell had a PEmpty
probability to be unsuitable for plant growth. To generate a regular pattern (Reg) for each density level, we used a more elaborate algorithm: when the two species densities were equally abundant (D= 50), the pattern was easily generated using a checkerboard
pattern with alternate occupation of cells by the two species. During this process, each cell had a probability of PEmpty to remain empty. This resulted in an alternating pattern
in which the direct neighbours (four nearest cells) of each cell were either empty or occu-pied by individuals of the other species. When the attractive and less attractive species' densities were not equal (D = 90 or D = 10), we divided each grid row into arrays of 10
cells. In each array, one cell was randomly assigned to either the less attractive species (D = 90) or to the attractive one (D = 10). The remaining cells were lled with the
other species or stayed empty with a probability PEmpty . We repeated this process ten
times in each row before lling the remaining rows in the same fashion. An example for the generated spatial patterns for the three density levels is shown in Figure 1.
Pollinators
The pollinator population represented one pollinator species with N individuals. The
pollinators could distinguish between the two ower species (unlike, for example, Straw 1972, Bobisud & Neuhaus 1975, Feldman et al. 2004) and were inherently more attracted to one of the plant species, i.e. at the beginning of each model generation, pollinators had a higher preference for the attractive species. During each generation, pollinators could change their preference according to their foraging experience (see section Pollinator foraging rules). While foraging, pollinators had a restricted detection range, called here Field Of View (FOV). The FOV was dened by a Moore neighbourhood with a radius of
RF OV cells (Wolfram 1983) which did not change across simulations, i.e. the number of
cells in the neighbourhood was equal to (2 Ö RF OV + 1)². We assumed that pollinators
about total reward distribution and there was no information exchange among pollinators (Pasquale & Jacobi 1998). We chose a pollinator population size that was smaller than the number of owers in the eld to ensure that there was a potential for competition among owers for pollinator visits (Straw 1972, Bobisud & Neuhaus 1975).
Pollinator foraging rules
To prevent articial aggregation of pollinators, each pollinator was randomly positioned in the plant community grid at the beginning of each model generation. From this random point, pollinators started their foraging bout using a specic decision rule. First, pollina-tors needed to decide about the rst ower sampled in the eld. This decision is usually assumed to be based on the inherent preferences of the pollinators. However, the initial choice may also be aected by a more general perception of the oral resources, because pollinators have a restricted ability to discern among oral resources at long distances (Ne'eman & Kevan 2001). Therefore, pollinators created a rescaled attractiveness land-scape of the owers in their eld of view (FOV), i.e. a landland-scape of attractiveness values which depended not only on the specic ower species in the cell, but also on the species growing in neighbour cells of increasing distances, as described in Equation 1:
A0ij = i+RF OV P k=i−RF OV k6=i j+RF OV P l=j−RF OV l6=j Akl d(ij),(kl) +Aij A0ij,max (1)
whereA0ij is the rescaled attractiveness of cell (i, j),Aij is the attractiveness of the ower
located in(i, j), d(ij),(kl) is the distance between a cell situated in(k, l) and the focal cell
in(i, j), andA0ij,max(RF OV) is the maximum value of rescaled attractiveness for the plant
at position(i, j)(i.e. if all cells in the neighbourhood are occupied by owers of maximum
attractiveness (A = 1)). It should be noted that A0ij,max(RF OV) is a function of the size
of the pollinators' FOV. Empty cells had a (rescaled) attractiveness of zero.
Once the rescaled attractiveness (A0ij) was obtained for all plant individuals, we let the
pollinators move from their randomly assigned positions in the direction of the steepest positive gradient of rescaled attractiveness within their specic FOV (i.e. in the direction of the more attractive part of the community). Pollinators followed that gradient until three conditions were met: (i) the chosen cell was not empty; (ii) no other pollinators were present in the chosen cell; and (iii) the numeric value of the steepest gradient was lower than a threshold value (T = 0.05). A threshold was chosen because otherwise, all
landscape (see Appendix 5). Therefore, the threshold parameter may be interpreted as a measure of the pickiness of pollinators concerning the starting point of their foraging bout. When two pollinators landed on the same grid cell, the later to arrive changed its starting random position by ying away for a distance of 25 cells in a randomly drawn direction (angle in [0,360°[) and started its search again until all three conditions were met.
After all pollinators chose their starting position, a second set of rules was applied. This second phase in the pollinator movements reected the common foraging mode of most pollinators, where short distances between consecutive visits are preferred (Waser 1982). This implies that pollinators were able to distinguish between dierent oral resources by detecting the individual attractiveness (Ne'eman & Kevan 2001). To create a short dis-tance foraging rule, pollinators searched for the most attractive ower within their FOV. The decision was made by each pollinator by calculating a score for all the cells within its FOV. This score was based on the distance between the location of the pollinator, the location of the plant individual and the instantaneous pollinator preference G (which was aected by its experience while foraging) as calculated in Equation 2:
S =G+ 1
(d+ 1) (2)
whereS is the instantaneous score for a certain plant individual in a specic cell,Gis the
instantaneous pollinator preference and d is the distance of the cell from the pollinator
(adapted from Ohashi & Thomson 2005).
The cell with the highest score value (S) in the pollinator's eld of view was chosen as
the pollinator's next destination, as long as it was not occupied by another pollinator, and it was not the pollinator's last visited ower in this generation. If a pollinator was already present in the chosen cell, the last arrived ew 25 cells away in a randomly chosen direction (angle in [0,360°[). If more than one ower within the FOV met the conditions and had an equal score, the next position was randomly chosen among these owers. To make sure that no bias was created during this step, pollinator order in this second phase was random.
Reward collection and pollination
Once the choice of the next ower was made (based on the second step of the pollinator foraging rules described above), the pollinator landed in that cell. As described in section Plant community, each plant individual had a PR probability to contain a reward. If the
Table 1: Shift in preferences of pollinators depending on the last species visited (rows) and the rewarding character of this last visit (columns). Numbers in brackets give the probability of the events to happen. PShif t is the constancy parameter of pollinators.
Rewarding character of last visit
Rewarding visit Non rewarding visit Last species attractive preference set to attractive: preference set to less attractive:
Stay (1) Shift (PShif t )
visited less attractive preference set to attractive: preference set to attractive: Stay (1-PShif t ) Shift (1)
pollen from the last visited ower and carried it only until the next visit. If the species of pollen carried matched the species of the currently visited ower, pollination occurred. If the species did not match, no pollination took place, but no additional negative eect was applied, because the limitation put on the pollen carryover time is akin to strong negative eects of heterospecic pollen transfer (Feldman et al. 2004). This foraging behaviour was repeated 50 times for each pollinator. Hence, each pollinator had the possibility of creating a maximum of 50 pollination events within one model generation (no pollen was carried by the pollinators at the beginning of a new vegetation generation).
Changes in pollination preferences during a bout
Pollinators were modelled as having an innate preference, dictating an inclination towards the attractive species at the beginning of each generation. However, within a genera-tion, the pollinator preferences could shift according to a set of decision rules (Table 1). Depending on the pollinator's constancy parameter (PShif t) used in our probabilistic
Win-Stay-Loose-Shift Behaviour algorithm (Ohashi & Thomson 2005), we could model a constant (i.e., specialist; PShif t = 0) as well as a shifting behaviour (i.e. generalist;
PShif t =1). Using this approach, the pollinator's past experience aected the
attrac-tiveness of species. Numerically, this was achieved by setting the pollinator preference values equal to the adequate plant attractiveness values for each individual pollinator. In this way, a pollinator that was not rewarded by an attractive species in a visit had
a PShif t probability of changing its preference towards the less attractive species in its
next visit. For this specic pollinator, the instantaneous pollinator preference G for the
less attractive species changed its value to equal the value of the attractive species and vice-versa.
Simulation experiments
To test the eect of pollinator characteristics on plant-pollinator interactions, we ran all the model simulations, i.e. all combinations of spatial patterns and density, using several pollinator trait combinations (see Table 2 for all the combinations of the parameter values used in the simulations). First, we tested the eects of pollinator constancy (PShif t) on the
dynamics of the model, tuning the pollinators from specialist behaviour (to the attractive species; PShif t = 0) toward a more realistic scenario in which pollinators presented a
shifting behaviour with innate preference towards the attractive species (PShif t = 1). To
do this, we considered 50 pollinators (N = 50) and used a priori dened values for the
other parameters (PR= 0.50,RF OV = 5 andddisp= 5). Because the changes in preferences
of pollinators were driven by the rewarding character of their visit (Keasar et al. 1996), we investigated the eect of promoting the pollinators' staying behaviour by increasing rewarding probability (PR= 1; i.e. each plant individual is rewarding at the start of each
generation and thus no reward variation occurred within species). In a further step, we aimed at exploring the interplay between the size of the monospecic vegetation patches and the maximum detection distance of pollinators. Hence, we conducted simulations in which the values for the pollinator's FOV were altered (RF OV = 2, 5 and 10). Finally,
to estimate the generality of our ndings, we conducted a sensitivity analysis for seed dispersal distances (ddisp = 2.5, 5 and 10) and pollinator population sizes (N= 25, 50 and
100). For each set of parameter values, 50 replications were made for each of the four starting patterns (Reg, S1, S5 and S10) and for the three relative density levels (D = 10,
50, 90).
Statistical analyses
The main goal of the statistical analysis was to estimate the probability of the less at-tractive species to survive in the community under dierent combinations of plant and pollinator characteristics (see Table 2 for all the combinations of the parameter values used in the simulations). Therefore, we mainly used survival analysis techniques, testing for dierences between Kaplan-Meyer estimates of the survival curves for the dierent spatial patterns within the same set of parameter values (Kleinbaum & Klein 2005). If data was censored, we used Log-Rank tests, whereas if no censoring was present, we used Mann-Whitney U tests (Kleinbaum & Klein 2005). As a rst step, we tested for dier-ences in survival between the four spatial patterns. If the appropriate test was signicant, we used the false discovery rate correction method to detect pairwise dierences (Ben-jamini & Hochberg 1995). All statistical analyses were conducted in R version 2.13.1 (R
Table 2: Parameter value combinations tested in the simulation experiments, wherePShif tis the
constancy parameter,RF OV is the Field Of View (pollinator detection distance, in cells),ddispthe
standard deviation of the seed dispersal distance,N is the size of the pollinator population and PR is the rewarding probability (i.e. reward variation) of each plant individual. All parameter
sets were run for all the combinations of four starting spatial patterns (S1, S5, S10 and Reg) and three densities (D= 10, 50 and 90).
Simulation N° Eect PShif t PR RF OV ddisp N
1 constancy high 0 0.5 5 5 50 2 constancy interm. 0.5 0.5 5 5 50 3 constancy low 1 0.5 5 5 50 4 no reward variation 0.5 1 5 5 50 5 no reward variation 1 1 5 5 50 6 FOV small 1 0.5 2 5 50 7 FOV large 1 0.5 10 5 50 8 dispersal short 1 0.5 5 2.5 50 9 dispersal long 1 0.5 5 10 50
10 polli. pop. small 1 0.5 5 5 25
11 polli. pop. large 1 0.5 5 5 100
Development Core Team 2005). The graphical results of the survival analysis of all the parameter combinations tested are presented in Appendix 2.
Results
Eect of density and spatial patterns of the plant communities
The spatial patterns and relative densities of the species composing the community sig-nicantly aected the survival rate of the less attractive species, even if eventually it went extinct in many simulations. Generally, the eect of spatial patterns on the survival rate of the less attractive species varied among relative densities: when the less attractive species was dominant at the beginning of a simulation (D= 10) it survived longer when spatially
aggregated (e.g. pattern S10). Vice-versa, at high density of the attractive species (D =
90), the less attractive species proted from being spatially dispersed (e.g. pattern S1; Table 3: simulations 3, 6, 8, 10, Figure 2: simulations 3, 6, Figure 3: simulations 8, 10 and Appendix 2 Figure A.1: simulations 3, 6, 8, 10). At intermediate densities, the results were similar to the eect found at high density of the attractive species, though weaker. In addition, the relative increase in density of the attractive species negatively aected
T able 3: Results of th e surviv al analysis for all para meter com binations. Eac h sim ulation exp erimen t w as de ned acc ording to N = num ber of po llina tors in the mo del, R F O V = size of polli nator Field Of View, P S hif t = sh iftin g parameter of the po llina tor beha viour (0: sp ecialist beha viour; 1:generalist beha viour), D = rel ativ e densit y of the attractiv e sp ecies, d disp = sta ndard deviation of the normal distribution used for mo delling seed disp ersal, P R = probabilit y of plan t ind iv iduals to be rew arding ). For eac h sim ulation exp erimen t, the median surviv al time (i.e. th e generation at whic h 50% of the sim ulations led to extinc ti on of the less attractiv e sp ecies) is giv en for th e fo ur spatial patterns: Reg regular pattern; S1, S5 and S10 clump ed spatial patterns with monosp ecic square patc hes with ed ge length of one, v e or ten ce lls, resp ectiv ely; the p-v alue of the adequate test (LR: Log Rank test; U: Mann Whitney U test) and the p-v alues of the pairwise comparison s (false disco very rate). V alues for p<0.05 are in italics, p<0.0 1 are und erlined, bold values indicate p<0.001. If less than 50 % of the sim ulations led to the extinction of the less attractiv e sp ecies, the median could not be computed, implying that the sp ecies surviv ed un til th e end of the sim ulation (indicated by >50). Median P airwise comparison N R F O V P S hif t D d disp P R Reg S1 S5 S10 test p-v alue surv ival S1-S5 S1 -S 10 S1-Reg S5-S10 S5-Reg S10-Reg 1 50 5 0 10 5 0.5 2 2